- Boost-users - lists.boost.org

[Range, Container] range of to-be-moved values ?
by John M. Dlugosz 05 Feb '14

05 Feb '14

Given a boost::container::vector<C> vec; I can write vec.push_back(move(value)) for a move-only (non-copyable) type C, instead of vec.push_back(value) as usual. But what would I write instead of vec.insert(pos,b,e) where (b,e] is a range of values? I'm particularly fond of Range concept, rather than loose iterators, so my function would be something that takes something that works as a Range; e.g. template <typename R> void prepend (const R& src_range) { auto b= boost::begin(src_range); auto e= boost::end(src_range); vec.insert(vec.begin(), b, e); } An rvalue reference to a range is not the same thing as a range of values that I may move from. I could have a differently named function and pass it a range with the implicit assumption that the function will modify the values, but is there a proper way to indicate that as part of the type? src_range can be, say, a pair of (non-const) iterator s as opposed to a pair of const_iterator s, or otherwise range_iterator<R>::type is dereferencable to a const value or not. But that itself is hard to see in general let alone overload against! If my function were called prepend_and_destroy_original(src) it might clue in the caller, but in general it is surprising that a parameter of (const R&) will modify the values. src itself doesn't change, and now refers to the same range of values which are now swapped out with defaults. With built-in pointers and even shared_ptr I can distinguish between an in-out parameter and a parameter that dereferences to a modifiable or const value. But what about with Ranges? Even with lower-case ranges provided by two iterators, How can you call something like boost::container::vector v2 (InIt_start, InIt_end); to convert a std::vector or some other container of move-only items? Only if the types match exactly can the vector(vector&&) constructor be used. —John

4 6

[iostreams] Why does boost::iostreams::copy close the Sink & Source?
by Sebastian Karlsson 05 Feb '14

05 Feb '14

What's the rationale behind this decision? I don't find it intuitive and reduces the utility of the function. In my case I read from a file into a stream which I want to later use, the signature of the device is important. But as soon as I've copied my data over to the device it's destroyed and along side the data goes down the drain. Kind regards, Sebastian Karlsson

1 0

Boost.MultiIndex. Using several keys together.
by syvyi 05 Feb '14

05 Feb '14

Hello, there's a structure struct person { std::string name; std::string surname; int birth_year; std::string craft; }; How to create such a multiindex container that dumping could performed using an index as name, surname and year triple alltogether? -- View this message in context: http://boost.2283326.n4.nabble.com/Boost-MultiIndex-Using-several-keys-toge… Sent from the Boost - Users mailing list archive at Nabble.com.

3 4

boost::python wrapping a container of boost::variant?
by Mario Lang 05 Feb '14

05 Feb '14

Hi. Does anyone have ideas how to convince boost::python to wrap a container of boost::variant? I need this to generate useful python bindings for a project of mine which uses boost::variant internally. Since python supports heterogenous containers natively, it seems to me this should be doable. However, I have attempted to find a solution on my own on several occasions, and always failed. One obvious hint is the indexing_suite. However, now that I looked at it again, I think I know where the problem is: The get_item method of indexing_suite always returns a data_type object, which is value_type from the container in question. However, from what I can see, I'd need something that converts to boost::python::object, so that I can return whatever class is appropriate for the actual variant instance. In other words, get_item would need to return a boost::python::object instead of data_type. To illustrate in code what I want, here is a minimal example: #include <boost/python.hpp> #include <boost/variant.hpp> struct a { int x; }; struct b { double y; }; typedef boost::variant<a, b> variant; typedef std::vector<variant> vector; BOOST_PYTHON_MODULE(bpv) { using namespace boost::python; class_<a>("a", init<a>()) .def(init<>()) .def_readonly("x", &a::x); class_("b", init()) .def(init<>()) .def_readonly("y", &b::y); } The question is, how do I expose std::vector<variant> to Python? I guess I need to define a static_visitor that returns python objects, something like: struct to_python_visitor : boost::static_visitor<boost::python::object> { result_type operator()(a const &a_) const { // return a boost::python::object that refers to a_ } ... }; And hook that up to the container support somehow such that boost::apply_visitor(to_python_visitor(), variant_instance) is called whenever a value is retrieved from the container. Can this even be done with current Boost.Python? Any hints, maybe even concrete code examples, would be highly appreciated. -- CYa, ⡍⠁⠗⠊⠕

1 0

boost::none for boost::shared_ptr
by Chris Stankevitz 03 Feb '14

03 Feb '14

Hello, Does boost::shared_ptr offer a way to create an empty shared pointer without a lot of typing? Perhaps something analogous to boost::none. I am not using c++11. Thank you, Chris == using boost::none; using boost::optional; // declare f(optional<CSomeReallyLongClassName> Item); g(shared_ptr<CSomeReallyLongClassName> pItem); // invoke f with empty object f(none); // invoke g with empty object g(shared_ptr<CSomeReallyLongClassName>());

5 9

multi_array move semantics
by MM 03 Feb '14

03 Feb '14

Hello, 1. Browsing the multi_array doc, there doesn't seem to be a move ctor? 2. Has there been any attempt to implement one? 3. if not, and the authors are reading this, are there any hints to help with this? Regards MM

2 1

serialization: intrusive_ptr
by MM 03 Feb '14

03 Feb '14

Hello, I've found an implementation of a serializer for intrusive_ptr here: http://stellar.cct.lsu.edu/files/hpx_0.9.5/html/code/hpx/util/serialize_int… which i've injected into my boost lib. Hartmut Kaiser, I guess, would there be a proposal to add it serialization for future standard boost? Regards MM

1 0

inter thread communication with message queues
by Kenneth Adam Miller 31 Jan '14

31 Jan '14

I was reading about message queues with respect to inter thread communication, and I have a specific use case that I want to describe. It's rather complex, so I want to be careful that I design it so that there are no race conditions. Pool A is a buffer (of type 1) handle manager, that feeds buffer handles into thread set 1, and receives old handles from thread set 2. There is only one thread running pool A. Thread set 1 is a set of N threads that fill up buffers of type 1 with regular data. It pushes full buffers to thread set 2. Thread set 2 is a set of M threads that compress buffers of type 1 into data buffers of type 2. Once the buffer of type 1 has be used up, it is cleared and sent back to Pool A. It pulls fresh data buffers of type 2 from Pool B, and sends freshly filled compressed buffers to Pool C. Pool C does something with the buffer, clears it and hands it to Pool B. Can this be done safely using an implementation like what I see here: http://www.justsoftwaresolutions.co.uk/threading/implementing-a-thread-safe… ? Note 1: -= denotes a 1-to-N connect and =- a N-to-1 connect RR = Request reply relationship +--------------------------------------------------------------------+ | | v | Pool A RR Thread Set 1 Router X | PULL-PUSH --==> PULL-read-PUSH ==--> PULL-PUSH --==> PULL Thread Set 2 PUSH \ / +-compress-+ Pool B RR / \ PULL-PUSH --==> PULL PUSH ^ || | || +----------- PUSH-write-PULL <--==++ Pool C Questions: how do I have a separate thread poll on multiple shared message queues? Like, I have a thread pool that both receives requests for buffer handles and that receives old buffer handles to recycle.

3 3

asio/ip/udp.hpp causes MSVC to use std::bind
by Chris Stankevitz 30 Jan '14

30 Jan '14

boost 1.55 VS2012 Update 4 === Can you tell me why the following code fails to compile? It smells as if udp.hpp is causing MSVC to "using namespace std". PS: It doesn't happen if I remove the asio include. PPS: It doesn't happen if g() takes int vs std::string. PPPS: It doesn't happen on GCC Thank you, Chris === #include <boost/bind.hpp> #include <boost/asio/ip/udp.hpp> #include <string> using boost::bind; void g(const std::string&) { } int main() { bind(g, _1); return 0; } === 1>------ Build started: Project: OverlayServer, Configuration: Debug Win32 ------ 1> main.cpp 1>main.cpp(13): error C2668: 'std::bind' : ambiguous call to overloaded function 1> c:\program files (x86)\microsoft visual studio 11.0\vc\include\functional(1387): could be 'std::_Bind<_Forced,_Ret,_Fun,_V0_t,_V1_t,_V2_t,_V3_t,_V4_t,_V5_t,<unnamed-symbol>> std::bind<void,const std::string&,boost::arg&>(_Rx (__cdecl *)(const std::basic_string<_Elem,_Traits,_Alloc>),_Vx0_t)' [found using argument-dependent lookup] 1> with 1> [ 1> _Forced=true, 1> _Ret=void, 1> _Fun=void (__cdecl *const )(const std::string &), 1> _V0_t=boost::arg<1> &, 1> _V1_t=std::_Nil, 1> _V2_t=std::_Nil, 1> _V3_t=std::_Nil, 1> _V4_t=std::_Nil, 1> _V5_t=std::_Nil, 1> <unnamed-symbol>=std::_Nil, 1> I=1, 1> _Rx=void, 1> _Elem=char, 1> _Traits=std::char_traits<char>, 1> _Alloc=std::allocator<char>, 1> _Vx0_t=boost::arg<1> & 1> ] 1> c:\externalpackages\boost_1_55_0\boost\bind\bind_cc.hpp(26): or 'boost::_bi::bind_t<R,F,L> boost::bind<void,const std::string&,boost::arg>(R (__cdecl *)(B1),A1)' 1> with 1> [ 1> R=void, 1> F=void (__cdecl *)(const std::string &), 1> L=boost::_bi::list1<boost::arg<1>>, 1> I=1, 1> B1=const std::string &, 1> A1=boost::arg<1> 1> ] 1> while trying to match the argument list '(overloaded-function, boost::arg)' 1> with 1> [ 1> I=1 1> ] ========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ==========

2 2

[accumulators] "count" NaN but otherwise ignore it?
by Rhys Ulerich 29 Jan '14

29 Jan '14

Is there a feature within accumulators to count NaNs on extraction but to otherwise ignore them? Maybe some sort of pass filter on incoming data? I've got a situation where I'd like to avoid NaNs clobbering otherwise useful information, but I still want to total count of data seen to include them. The context is a time integrator for a physics simulation. The integrator tracks the number of timesteps. When the physics kicks back a NaN (because something's amiss), I'd like to count the failed timestep but not have the NaN destroy the useful mean of timesteps taken thus far. Thanks, Rhys

2 1